3.4 — Explain It Back | AI for Real Life

Explaining it out loud is how you find your gaps.

The biggest difference between "I feel like I understand this" and "I can actually use this on a test" is whether you've tried explaining it out loud, in your own words, to someone who's allowed to ask follow-up questions. Explaining forces your brain to organize the material: and surfaces every fuzzy spot you'd otherwise miss until the test.

The hard part used to be finding someone willing to listen at 11pm on a Wednesday. Now you have Claude, who is never bored and asks better follow-ups than your roommate. Five minutes of explain-it-back is worth 30 minutes of re-reading.

The flip: Claude becomes the student.

Role reversal. Claude is normally the expert. Here: Claude is a curious-but-clueless student. You teach. Claude (as the student) asks the follow-ups that find what you don't actually know.

The Explain-It-Back prompt: the signature move

We're going to flip roles. You're a smart but curious student who has NOT studied [topic: be specific]. I'm going to teach it to you in my own words. Your job: 1. Listen to my explanation. 2. After I'm done, ask me 3 follow-up questions: the kind a smart, slightly-skeptical student would ask. Find the parts where my explanation was vague, where I used a term without defining it, where I skipped a step, or where I oversimplified. 3. Don't be polite. The whole point is to surface what I don't actually understand. 4. After I answer your follow-ups, give me a brutally honest assessment: - What I clearly understand well - What I'm fuzzy on (and how I can tell) - What I got wrong - The ONE concept I should re-study before doing this again Don't grade me on a curve. Don't tell me I did great if I didn't. I'd rather find the gap now than on the test. Topic I'll be teaching: [topic] Class context: [class name + level: so you know how deep to push] I'll start teaching now. Stay quiet until I say "your turn."

Out loud beats typing: use voice mode

Voice mode (Lesson 1.3) was built for this. Walking back from the dining hall, headphones in, explaining the Krebs cycle out loud to Claude. It looks weird. It works well: the act of speaking the explanation surfaces gaps faster than typing it.

The three failure modes (and what each one means).

Run the prompt enough times and you'll notice your explanations fail in three distinct ways. Each one tells you something different about what to do next.

✅

Pattern 1: clean teach, easy follow-ups

What it means: you actually know this. Move on.
Don't: re-study this topic: that's spending time on the wrong thing.
Do: move to the next topic on the study guide and run this prompt again.

⚠️

Pattern 2: okay teach, one follow-up exposes the gap

What it means: you half-know it.
Do: go back to the source on JUST the gap (re-read that paragraph, re-watch the relevant slice of the lecture). Then re-explain.

🛑

Pattern 3: can't even start a clear sentence

What it means: you don't know this yet. You thought you might (recognition tricked you).
Do: pause explain-it-back. Re-learn from the source: lecture, chapter, slides. Then come back and try again.
Don't: ask Claude to "just explain it to me." That feels productive but puts you back in recognition mode without ever building the generation skill.

The cleanest example of Rule 1 in the whole course

Honest Work Code · Rule 1: Learn with it, not instead of it. Explain-it-back is the highest-honesty AI use for school. Claude isn't doing the work. Claude isn't giving you the answer. Claude is doing one thing humans rarely have time, patience, or interest to do: listening to your explanation and asking the question that surfaces your blind spot. It's a move you can use on graded material with zero integrity worry, in any class, under any policy, because Claude wrote nothing. You did all the talking.

Two variations to keep in your back pocket.

Once the basic prompt is muscle memory, two variations are worth saving. The first stress-tests for depth. The second is for any pair of concepts you keep mixing up.

Variation 1: The "explain it three ways" calibration

I want to test whether I REALLY understand [topic]. Make me explain it three different ways, in this order: 1. To a 5th grader, in two sentences, no jargon at all. 2. To a freshman in this class, in a paragraph, with one example. 3. To a peer in office hours, with the technical precision the prof would want. After each one, ask me a follow-up question that would only land if I actually understand. Then tell me which version was strongest and which was weakest: that's where the real gap is.

Variation 2: The "compare and contrast" trap

Explain to me the difference between [Concept A] and [Concept B]:two things I keep mixing up in [class]. Listen to my explanation. Then: 1. Ask me to give you an example of A that is NOT also true of B (and vice versa). 2. If I can't, that's the gap: drill in until I can. 3. Once I've nailed the distinction, ask me one tricky case where it would be ambiguous which one applies. Make me reason through it out loud. Be picky. Two concepts that sound similar are exactly where multiple-choice traps live.

Try this: explain something out loud, right now, before the next lesson

Pick a topic you have a quiz or test on this week. Open Claude. Switch to voice mode if you can.

1. Run the basic Explain-It-Back prompt.
2. Teach the topic out loud. Don't peek at notes.
3. Answer the follow-ups honestly.
4. Write down the ONE concept Claude said you should re-study. That's tomorrow's first study task.

Up next: the lesson that uses everything you've built so far.

You have clean lecture notes (3.1), unified study docs (3.2), an active recall engine (3.3), and explain-it-back (3.4). Lesson 3.5 is where they all click together: the 4-Day, 2-Day, Day-Of midterm stack.

Continue to 3.5 → The Midterm Stack